| In this thesis we present "END", Expert Network Designer, an automated system for large structured computer networks design, modeling, simulation and performance evaluation.; END employs formalized network design experience to recommend the feasible network designs suitable for the particular user's network environment, and explain why such solutions have been chosen. END further ranks and evaluates such solutions using artificial intelligence optimization techniques and engineering simulation tools.; END has been equipped with a neural network learning utility to improve the time-efficiency of its network design problem solver, and a machine learning utility to enable the system to learn about new emerging technologies, their optimal design techniques, and the evolution of the specifications of the existing technologies.; Also in this thesis a fuzzy expert system approach is proposed as a replacement for the classical expert system to alleviate some restrictions with the classical hard-decisions expert system approach. Finally some ideas of applying hybrid fuzzy expert systems are discussed.; By employing such integrated subsystems, we are able to obtain (1) network design solutions suitable for the user's network environment, (2) reasoning analysis of why these specific solutions have been chosen, (3) performance evaluation for the different design options, and (4) analytical ranking of the proposed solutions. |
